What happened
An opinion piece examines Andrej Karpathy's optimistic claim that AI democratizes knowledge by providing equal access to LLMs regardless of wealth, but argues the reality is more complex—history suggests major technological shifts concentrate power among the few rather than distributing it evenly.
Why it matters
While LLMs offer cheap, instant expertise on paper, barriers such as $200/month subscription costs, lack of domain expertise, and unequal ability to leverage AI constructively mean the technology risks amplifying inequality rather than reducing it. Students using AI to skip deep learning face a higher job market bar, while industry experts with proprietary data gain disproportionate advantage—a pattern seen with the Internet, which promised democratization but ended up concentrating gatekeeping power among multinational corporations.
What to watch
The author argues that open-weight models, local inference, and an open-source AI ecosystem are essential to prevent total enclosure and foster competition—drawing a parallel to how open-source software shaped the Internet, and calling this approach "the surest way to bring power to the people."
Summaries like this, in your inbox every morning.
The piece frames AI as the latest technological inflection point that carries both utopian and dystopian potential, drawing on Andrej Karpathy's tweet asserting that LLMs democratize expertise by offering the same model to billionaire and individual alike. However, the author argues that equal access on paper does not translate to equal power in practice. The historical parallel to the Internet is instructive: that network was envisioned as a decentralized library of human knowledge, yet it became dominated by a handful of multinational corporations that control most digital gateways and have weaponized the platform for surveillance and censorship. The author identifies three concrete mechanisms by which AI may replicate and amplify this inequality: economic (subscription costs exclude most of the world), epistemic (domain expertise and proprietary data give industry insiders a "data flywheel" advantage), and behavioral (unequal critical capacity to use AI constructively versus delegating thought entirely). The piece cites an emerging labor-market symptom: students who leaned on AI shortcuts during education now face steeper hiring bars precisely because employers anticipate workforce compression and are eliminating entry-level roles. The author positions open-weight models, local inference, and an open-source ecosystem as structural countermeasures, invoking the success of open-source software in shaping the Internet's trajectory. The call is urgent: without deliberate action to preserve openness and competition, AI risks concentrating power further rather than distributing it.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
Washington revived its AI and defense infrastructure push, with plans for an "AI Force" and an AI tsar
China's spy agency chief warned last week that AI's hacking abilities could jeopardize national security, and…
The New York Times and other news organizations filed new court documents citing OpenAI and Microsoft's intern…
Abeam Consulting and Notion are promoting an effort to shift companies to AI-driven operations and organizatio…
Zscaler introduced "Zscaler Agentic SOC," which embeds AI agents into security operations to support detection…
Generative Partners began offering "AX BPO" in September 2026, a BPO service that handles exceptions, visual c…
